Outcome Prediction in Intensive Care Units Based on Clinicial Adverse Events: A Clustering Model Approach

Á. Silva, R. Martins, M. Santos, L. Gomes, and J. Neves (Portugal)


Mortality Predicting Models based onIntermediate Outcomes, Knowledge Discovery fromDatabases, Data Mining, Databases, Artificial NeuronalNetworks, Artificial Intelligence.


During the past 20 years, Intensive Care Units (ICU) risk prediction models have undergone significant development, validation and refinement. Among the general ICU severity scoring systems the Simplified Acute Physiology Score (SAPS II) [l] have become the most accepted and used. Recent innovation risk adjustment includes the use of Artificial Intelligence (AI) techniques and Knowledge Discovery from Database (KDD) where the clinical data of patients are collected in large Datadases (DB). These innovations have the potential of extending the uses of case severity-of-illness adjustment in areas of clinical outcome research and patient care. The objective of this work is to demonstrate that the SAPS II score, which allows the estimation of the patient probability of death in hospital, can be complemented or even substituted by the modern AI techniques. We reach to prove that Clinical Adverse Events (CAE) based approaches are better in evaluation of risk mortality as a complement of current SAPS II. Applying clustering techniques we can trace a patient's evolutionary line (related to the process of care), by the analysis of the intermediate outcomes.

Important Links:

Go Back